Direct current support capacitor capacitance value online identification method based on multi-frequency point collaborative analysis

Abstract

The invention provides a direct-current support capacitor capacitance online identification method based on multi-frequency-point collaborative analysis, and relates to the technical field of direct-current support capacitors, and the method comprises the steps: collecting a ripple voltage signal and a ripple current signal; determining a delay step length, and constructing a pulse weight sequence based on the delay step length; performing discrete convolution operation on the pulse weight sequence, the ripple voltage signal and the ripple current signal, and extracting a periodic signal; performing signal processing on the periodic signal, and extracting a signal amplitude; determining an optimal parameter combination; an optimized amplitude value is obtained; calculating a first capacitance value and a second capacitance value according to the amplitude relation; and performing multi-frequency-point collaborative analysis based on the first capacitance value and the second capacitance value, and determining an actual capacitance value of the DC support capacitor. According to the method, high-precision extraction is carried out on signals of different frequency points through adaptive time slot pulse weighting, synchronous demodulation and integral filtering processing, and rapid and accurate identification of capacitance parameters of the direct-current support capacitor is realized in combination with a multi-frequency-point collaborative analysis mechanism.

Description

Technical Field

[0001] The present invention relates to the technical field of DC support capacitors, and in particular to an online capacitance identification method for a DC support capacitor based on multi-frequency collaborative analysis. Background Art

[0002] As key energy storage components in power electronics, DC link capacitors (DC link capacitors) perform core functions in power transmission and transformation systems, including reactive power compensation, harmonic suppression, DC bus voltage regulation, and high-frequency energy transfer. In high-voltage direct current (HVDC) converter stations, DC link capacitors maintain grid voltage stability through transient energy exchange. In flexible AC transmission devices such as static VAR compensators (SVCs) and static synchronous compensators (STATCOMs), DC link capacitors filter characteristic subharmonics. In multilevel power converters, DC link capacitors provide a low-impedance energy buffer, effectively suppressing DC voltage fluctuations caused by sudden power surges. According to statistics from the International Council on Large Electric Systems (CIGRE), capacitor-related failures account for 32% of transmission and transformation system faults, with repair costs approximately 5-8 times that of preventive maintenance. Furthermore, the capacitor bank capacity in UHV converter stations accounts for 15%-20% of the total equipment investment, and its reliability directly impacts system economics and power supply stability.

[0003] Degradation of DC link capacitor parameters poses a serious risk to the safe operation of power transmission and transformation systems. Capacitance degradation is a common failure mode of DC link capacitors. Under the long-term effects of high temperature, high current, and alternating electrical stress, internal electrode polarization and dielectric aging in the DC link capacitor lead to a reduction in effective capacitance area, which in turn causes a drop in capacitance. Film capacitors are commonly used as DC link capacitors. Their low equivalent series resistance, high withstand voltage, high ripple current capability, and long service life make them particularly suitable for DC link applications in high-voltage, high-power power electronics. When the DC link capacitor capacitance drops below 80% of the nominal value, an early warning is required; if it drops below 70%, it should be replaced immediately. Abnormal capacitance not only affects the dynamic performance of the equipment but also triggers a series of chain reactions: In DC transmission systems, insufficient DC link capacitor capacitance can lead to increased DC bus voltage fluctuations and increased converter output harmonics; in electric vehicle charging stations, it can cause excessive output voltage ripple and increase battery thermal stress; and in smart grids, it can degrade the response characteristics of power electronics, reduce system stability margins, and in severe cases, even trigger system protection and cause shutdowns.

[0004] Traditional DC link capacitor capacitance monitoring technologies primarily include impedance analysis, pulse response, and spectrum analysis. Impedance analysis calculates capacitance by measuring the impedance characteristics of DC link capacitors at specific frequencies, but this method is significantly affected by external electromagnetic interference. The pulse response method, based on the excitation-response principle, is simple to operate and low-cost, but suffers from a long dynamic response time (>800ms) and poses potential safety risks in high-voltage scenarios. Spectrum analysis supports non-contact measurement but is sensitive to background harmonics and requires complex digital signal processing algorithms. A systematic comparative study shows that existing methods exhibit an average error of 4.2%-6.8% for online monitoring in complex electromagnetic environments, with response delays exceeding 500ms. These methods fail to meet the accuracy and speed requirements for real-time monitoring of DC link capacitor capacitance parameters in new-generation power electronic equipment. Therefore, a high-precision, highly anti-interference, and fast-response online identification technology for DC link capacitor capacitance is urgently needed. Summary of the Invention

[0005] In view of this, the present invention proposes an online identification method for the capacitance of a DC support capacitor based on multi-frequency collaborative analysis. Through adaptive time slot pulse weighting, synchronous demodulation and integral filtering processing, high-precision extraction of signals at different frequency points is implemented, and combined with the multi-frequency collaborative analysis mechanism, rapid and accurate identification of the capacitance parameters of the DC support capacitor is achieved.

[0006] The technical solution of the present invention is achieved as follows:

[0007] The present invention provides an online identification method for the capacitance of a DC support capacitor based on multi-frequency collaborative analysis, comprising:

[0008] S1. Using a sensor system to collect ripple voltage signals and ripple current signals at both ends of the DC link capacitor;

[0009] S2. Determine a delay step length according to the grid fundamental frequency and the sampling frequency, and construct a pulse weight sequence based on the delay step length;

[0010] S3. Performing a discrete convolution operation on the pulse weight sequence, the ripple voltage signal, and the ripple current signal to extract a periodic signal corresponding to the grid fundamental frequency and its harmonic components;

[0011] S4. Process the periodic signal and extract the signal amplitudes at the grid fundamental frequency point and harmonic frequency point respectively;

[0012] S5. Traverse the optimization parameters through the parameter search algorithm to determine the optimal parameter combination;

[0013] S6. Re-execute steps S2 to S4 using the optimal parameter combination to obtain optimized amplitudes of the ripple voltage signal and the ripple current signal at the grid fundamental frequency point and harmonic frequency point;

[0014] S7. Based on the voltage-current relationship of the capacitor, calculate a first capacitance value according to the amplitude relationship of the optimized ripple voltage signal and the ripple current signal at the grid fundamental frequency point, and calculate a second capacitance value according to the amplitude relationship of the optimized ripple voltage signal and the ripple current signal at the harmonic frequency point;

[0015] S8. Perform multi-frequency collaborative analysis based on the first capacitance and the second capacitance to determine the actual capacitance of the DC link capacitor.

[0016] Preferably, in step S2, the delay step length D is determined by:

[0017]

[0018] where f s is the sampling frequency, and f0 is the fundamental frequency.

[0019] Preferably, in step S2, the construction formula of the pulse weight sequence h(n) is:

[0020]

[0021] Where δ(n) is the discrete unit pulse, M is the number of pulses, D is the delay step, and n is the discrete time index.

[0022] Preferably, step S1 specifically includes:

[0023] The ripple voltage signal and the ripple current signal at both ends of the DC link capacitor are collected at a fixed sampling frequency;

[0024] The data of the first N sampling points are selected for processing, where N is a preset value, to ensure that the selected data covers one or more complete cycles of the grid fundamental frequency.

[0025] Preferably, in step S3, the discrete convolution operation performs time-domain convolution on the pulse weight sequence with the ripple voltage signal and the ripple current signal respectively, thereby achieving selective enhancement and noise suppression of specific frequency components, and obtaining weighted voltage signals and current signals.

[0026] Preferably, in step S4, the signal processing includes multiplying the extracted periodic signal with the cosine function and the sine function of the corresponding frequency to obtain an in-phase component and an orthogonal component, and performing discrete integral filtering, amplitude compensation and tail-end averaging processing on the in-phase component and the orthogonal component.

[0027] Preferably, the multiplication of the extracted periodic signal with the cosine function and sine function of the corresponding frequency respectively means: for the grid fundamental frequency point and harmonic frequency point, the convolved voltage signal is multiplied with the cosine function of the corresponding frequency to obtain the voltage in-phase component, and multiplied with the sine function of the corresponding frequency to obtain the voltage orthogonal component; the convolved current signal is multiplied with the cosine function of the corresponding frequency to obtain the current in-phase component, and multiplied with the sine function of the corresponding frequency to obtain the current orthogonal component.

[0028] Preferably, the discrete integral filtering process includes:

[0029] Construct an integrating filter whose impulse response is of length L lp The uniform sequence, L lp is the integration window length;

[0030] Performing convolution operations on the integral filter with the in-phase component and the quadrature component of the voltage and the current, respectively, to obtain filtered in-phase component and quadrature component of the voltage and the current;

[0031] The amplitude compensation includes:

[0032] Amplify and compensate the filtered in-phase component and quadrature component to correct the amplitude attenuation caused by integral filtering;

[0033] The tail segment averaging process includes:

[0034] The voltage and current amplitudes at each time point are calculated separately, and the tail section of the signal is selected for averaging to obtain the final voltage and current amplitude estimates.

[0035] Preferably, in step S5, the search space of the parameter search algorithm includes: the value range of the number of pulses is 2 to 10, the value range of the integral filter length parameter is 1 to 5, and the value range of the tail segment average ratio parameter is 2 to 10.

[0036] Preferably, the multi-frequency collaborative analysis in step S8 includes:

[0037] When the difference between the first capacitance and the second capacitance is less than a preset deviation threshold, the arithmetic mean is used as the actual capacitance of the DC link capacitor;

[0038] When the difference between the first capacitance and the second capacitance is greater than or equal to a preset deviation threshold, preferentially selecting one of the capacitances as the actual capacitance of the DC link capacitor according to the signal-to-noise ratio;

[0039] The signal-to-noise ratio is determined by signal amplitude, spectrum characteristics or historical data statistical characteristics.

[0040] The present invention has the following beneficial effects compared to the prior art:

[0041] (1) The online identification method of the DC support capacitor capacitance parameters based on multi-frequency collaborative analysis proposed in the present invention realizes high-precision online identification of the DC support capacitor capacitance by collecting the ripple voltage and current signals at both ends of the DC support capacitor and combining pulse weight sequence construction, discrete convolution operation, synchronous demodulation processing and multi-frequency collaborative analysis technology. This effectively solves the technical problems of existing capacitance monitoring methods such as insufficient accuracy and delayed response in complex electromagnetic environments.

[0042] (2) The present invention adopts a signal processing method that combines adaptive time-slot pulse weighting with synchronous demodulation, so that the effective extraction of target frequency components is not affected by grid frequency fluctuations and noise interference. By performing discrete convolution operations on the collected ripple signal and the precisely constructed pulse weight sequence, and then combining orthogonal modulation decomposition and integral filtering processing, the accuracy of amplitude extraction is significantly improved, and the capacitance parameter identification error is controlled below 1%. Compared with the 4.2%-6.8% error rate of the traditional spectrum analysis method, it is significantly improved, meeting the technical requirements of the power system for high-precision monitoring;

[0043] (3) The present invention simplifies the computational process based on discrete integration and pulse weighting, reducing algorithm complexity and optimizing computational resource usage. Key parameters such as the integral filter length and tail segment average ratio are self-optimized through a parameter search algorithm, reducing the amount of data processing while ensuring measurement accuracy. This enables the system to achieve a response time of 30-50ms on an embedded platform, meeting the power system's technical requirements for real-time monitoring of capacitor status and effectively supporting rapid early warning and disposal of power grid faults.

[0044] (4) The present invention adopts a multi-frequency collaborative analysis strategy, processes the grid fundamental frequency and its harmonic frequency components at the same time, obtains the capacitance value at different frequency points, and determines the final capacitance value by setting a threshold comparison and signal-to-noise ratio analysis. This multi-frequency data fusion processing method effectively suppresses random interference and system errors in the power system, and improves the reliability and stability of capacitance calculation. Especially in an environment with high grid harmonic content, compared with the single-frequency analysis method, the multi-frequency collaborative analysis can more accurately reflect the actual state of the capacitor and provide a more accurate parameter basis for capacitor life assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0046] Figure 1 is a flow chart of the method of the present invention;

[0047] Figure 2 This is a technical implementation diagram of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0049] like Figure 1 As shown, the present invention provides an online identification method for the capacitance of a DC support capacitor based on multi-frequency point collaborative analysis, comprising:

[0050] S1. Using a sensor system to collect ripple voltage signals and ripple current signals at both ends of the DC link capacitor;

[0051] S2. Determine a delay step length according to the grid fundamental frequency and the sampling frequency, and construct a pulse weight sequence based on the delay step length;

[0052] S3. Performing a discrete convolution operation on the pulse weight sequence, the ripple voltage signal, and the ripple current signal to extract a periodic signal corresponding to the grid fundamental frequency and its harmonic components;

[0053] S4. Process the periodic signal and extract the signal amplitudes at the grid fundamental frequency point and harmonic frequency point respectively;

[0054] S5. Traverse the optimization parameters through the parameter search algorithm to determine the optimal parameter combination;

[0055] S6. Re-execute steps S2 to S4 using the optimal parameter combination to obtain optimized amplitudes of the ripple voltage signal and the ripple current signal at the grid fundamental frequency point and harmonic frequency point;

[0056] S7. Based on the voltage-current relationship of the capacitor, calculate a first capacitance value according to the amplitude relationship of the optimized ripple voltage signal and the ripple current signal at the grid fundamental frequency point, and calculate a second capacitance value according to the amplitude relationship of the optimized ripple voltage signal and the ripple current signal at the harmonic frequency point;

[0057] S8. Perform multi-frequency collaborative analysis based on the first capacitance and the second capacitance to determine the actual capacitance of the DC link capacitor.

[0058] like Figure 2As shown, the present invention provides an online identification method for the capacitance parameters of DC support capacitors based on multi-frequency collaborative analysis. This method collects the ripple voltage and current signals at both ends of the DC support capacitor, combines pulse weight sequence construction, discrete convolution operation and orthogonal synchronous demodulation processing, thereby achieving high-precision online identification of the capacitor capacitance. The technical implementation process of the present invention mainly includes the following five steps: A. Data acquisition and preprocessing, B. Target time pulse weighted processing, C. Orthogonal synchronous demodulation and integral filtering, D. Automatic search and optimization, E. Real-time capacitance calculation. These five steps will be explained in detail below.

[0059] A. Data Collection and Preprocessing

[0060] First, a sensor system deployed around the DC link capacitor is used to synchronously collect the ripple voltage signal v(t) across the DC link capacitor and the ripple current signal i(t) flowing through the capacitor. The collected analog signals are converted into digital signal sequences v(n) and i(n) by a high-speed analog-to-digital converter, where n represents the sampling sequence number. The system uses a fixed sampling frequency f s The analog signal is discretely sampled and the first N sampling points are selected for subsequent processing to ensure that the data covers one or more complete cycles of the grid base frequency. In actual implementation, the sampling frequency f s It is usually set between 10kHz and 50kHz to meet the sampling requirements of the grid fundamental frequency f0 (such as 50Hz or 60Hz) and its harmonic components. The system performs data window interception on the collected digital signal sequence, and the window length is usually set to f s / f0 is an integer multiple of , in order to include the complete periodic signal. At the same time, the system uses a high-pass filter to remove DC components and low-frequency interference, and a low-pass filter to suppress high-frequency noise, thereby improving the signal quality of subsequent processing. According to the fundamental frequency characteristics of the power grid, the delay step length D is calculated for the construction of the subsequent time slot pulse sequence. Its calculation formula is:

[0061]

[0062] where f s is the sampling frequency, and f0 is the fundamental frequency. Taking a 10kHz sampling frequency as an example, when the fundamental frequency is 50Hz, the delay step size D = 200, which means that every 200 sampling points corresponds to one fundamental frequency cycle.

[0063] B. Adaptive time-slot pulse weighted filtering

[0064] This step constructs a pulse sequence to extract the signal components synchronized with the target frequency, effectively weighting the original signal. The discrete convolution operation enhances the portion of the original signal corresponding to the target period while suppressing other noise and interference. The processed signal has a higher signal-to-noise ratio at the target frequency, providing clearer signal characteristics for subsequent orthogonal synchronous demodulation. The specific implementation steps are as follows:

[0065] B1. Constructing a time-slot pulse weighted sequence

[0066] Based on the delay step D calculated above, a pulse weight sequence h(n) is constructed. This sequence is a set of periodically distributed unit pulses, defined as:

[0067]

[0068] Where δ(n) represents the unit pulse function. When n = 0, δ(n) = 1; otherwise, δ(n) = 0. M is the number of pulses, an adjustable parameter ranging from 2 to 10, indicating how many cycles of data are used for processing. m is the pulse sequence number, ranging from 0 to M-1. n is the discrete time index. This sequence "focuses" on the target frequency component in the time domain, effectively suppressing other frequency components and noise interference during the subsequent synchronous demodulation process.

[0069] B2. Discrete Convolution Operation

[0070] Perform discrete convolution on the input signal v(n) and the pulse sequence h(n):

[0071]

[0072] Where v(n) is the input signal, h(n) is the pulse weight sequence, and v w (n) is the output signal after convolution, and n is the discrete time index.

[0073] In the frequency domain, the Fourier transform of a pulse sequence is a periodically repeating function. Convolution is equivalent to applying a sampling window function to the signal, selectively enhancing specific frequency components in the spectrum. Specifically, only when n aligns with the pulse position does the target periodic component in the signal receive cumulative enhancement; non-periodic or other frequency components, due to phase differences, cancel each other out, achieving noise suppression.

[0074] Similarly, the same processing is performed on the input signal i(n), and we get:

[0075]

[0076] C. Quadrature Synchronous Demodulation and Integral Filtering

[0077] In this step, orthogonal modulation technology is used to decompose the signal into orthogonal components, and discrete integral filtering is used to perform low-pass smoothing to remove high-frequency noise. Amplitude compensation and tail-end averaging strategies are used to ensure the extracted signal amplitude is stable and accurate. The processed amplitude data provides a key basis for error analysis and capacitor value calculation. Overall, this step ensures the integrity and accuracy of the signal information, which directly affects the measurement accuracy of the system. Specific implementation includes:

[0078] C1. Using cosine and sine waves that are orthogonal to the target signal as references, the target frequency component is mapped to two orthogonal components, thereby achieving amplitude and phase separation. The formula is described as follows:

[0079]

[0080] Among them, I(n) is the in-phase component, Q(n) is the quadrature component, and f target is the target frequency, f s is the sampling frequency and n is the discrete time index.

[0081] Similarly, the current signal is processed in the same way:

[0082]

[0083] To calculate the in-phase component I of the current signal i (n) and the quadrature component Q i (n).

[0084] C2. In order to eliminate the high-frequency interference and noise caused by the modulation process, a discrete integral filter is used. The filter impulse response is:

[0085]

[0086] Among them L lp is the integration window length, and L lp =length_lp×D, length_lp is the integral filter length parameter, D is the delay step size, and n is the discrete time index;

[0087] Perform convolution operations on the in-phase component and the orthogonal component separately:

[0088]

[0089] Among them, I lp (n) is the in-phase component after filtering, Q lp (n) is the orthogonal component after filtering; it is equivalent to performing a moving average on the signal to smooth out high-frequency noise.

[0090] Similarly, the same processing is performed on the in-phase component and quadrature component of the current signal:

[0091]

[0092] C3. After synchronous demodulation, due to the energy distribution problem of sine and cosine modulation (twice the signal energy is dispersed into two components), it is necessary to multiply by a compensation factor of 2:

[0093] I comp (n)=2I lp (n),Q comp (n) = 2Q lp (n)

[0094]

[0095] Get the amplitude estimate of the post-modulation compensation:

[0096]

[0097] Among them, A v (n), A i (n) is the amplitude estimation of voltage signal and current signal at each time point;

[0098] C4. To reduce the impact of initial transients or boundary effects, select the tail segment of the signal for averaging:

[0099]

[0100] in, is the final estimated signal amplitude of the voltage signal and current signal, is the starting point of the tail segment, N is the total number of sampling points; mid is the parameter that controls the average ratio; N tail =N-n0 is the number of sample points in the tail section used for averaging calculation.

[0101] Through the above processing, the voltage amplitude at the grid fundamental frequency point f0 can be obtained: and current amplitude And the voltage amplitude at the harmonic frequency point 2f0 and current amplitude

[0102] D. Automatic search and optimization

[0103] In this step, the system optimizes parameters for both voltage and current amplitude extraction to improve the accuracy of amplitude estimation. Because voltage and current signals have different characteristics (such as signal-to-noise ratio and harmonic content), different optimization parameter combinations may be required for each signal.

[0104] Specific implementation includes:

[0105] D1. The search parameters include the number of pulses M, the integral filter length parameter length_lp, and the tail average ratio parameter mid. For each set of parameters in the preset parameter space, the amplitude A at the target frequency of each channel is calculated according to the above steps. est .

[0106] Among them, the value range of the pulse number M is 2 to 10, the value range of the integral filter length parameter legnth_lp is 1 to 5, and the value range of the tail segment average ratio parameter is 2 to 10.

[0107] The optimization process uses a parameter grid search method, which traverses each parameter combination in the preset parameter space, calculates the amplitude estimate of each channel at the target frequency according to the above steps, and then compares it with the known reference value to calculate the error. The process is as follows:

[0108] Initialize the minimum error to a larger value;

[0109] For each parameter combination (M, length_lp, mid) in the parameter space:

[0110] Using this set of parameters, process the signal according to steps B and C to obtain the amplitude estimate A est ;

[0111] Calculate the relative error for each signal:

[0112]

[0113] Among them, A true is the pre-calibrated reference amplitude, A est is the calculated amplitude estimate;

[0114] If the relative error is less than the minimum error, the minimum error is updated to the current relative error, and the current parameter combination is recorded as the optimal combination;

[0115] Output the optimal parameter combination (M, length_lp, mid).

[0116] The optimization process is then repeated for both the voltage and current signals to obtain their respective optimal parameter combinations. These parameters are then used in actual measurements to ensure high-precision amplitude extraction.

[0117] D2. In addition to the grid search method, the system can also implement an adaptive optimization mechanism. During actual operation, the system continuously adjusts parameters through error feedback, gradually improving the accuracy of amplitude extraction. The error feedback mechanism is as follows:

[0118] 1. Set the error threshold ε th ;

[0119] 2. After each measurement, compare the current amplitude estimate with historical data or expected values;

[0120] 3. If the error exceeds the threshold, a local search is performed near the current parameters to find better parameters;

[0121] 4. Parameter update adopts weighted average method:

[0122] P new =αP old +(1-α)P better

[0123] Among them, α is the smoothing factor, ranging from 0.7 to 0.9, P better Better parameters found for local search.

[0124] This adaptive optimization mechanism enables the system to continuously learn and adapt to changes in signal characteristics during operation, thereby improving the robustness and adaptability of the system.

[0125] E. Real-time calculation of capacitance

[0126] In this step, the real-time calculation results are compared with the reference amplitudes obtained through pre-calibrated experiments to calculate the error ratio. Based on the error, processing parameters are further adjusted to achieve automatic correction, ensuring that the error of each frequency amplitude in the final output is below the preset threshold. Based on the voltage and current amplitudes obtained in the previous steps, the system calculates the capacitance of the DC link capacitor. The specific implementation process is as follows:

[0127] E1. First, according to the basic principle of capacitors, the relationship between capacitor current and voltage is:

[0128]

[0129] Where i(t) is the current signal flowing through the capacitor, C is the capacitance of the capacitor, is the rate of change of voltage across the capacitor.

[0130] E2. When the DC link capacitor operates with a sinusoidal ripple, the voltage signal is:

[0131] v(t)=V ripple sin(2πft)

[0132] Among them, V ripple is the amplitude of the voltage ripple, and f is the signal frequency.

[0133] Taking the derivative of the voltage signal, we get:

[0134]

[0135] Substituting the capacitor's current-voltage relationship into the equation, we get the current expression:

[0136] i(t)=C·2πf·V ripple cos(2πft)

[0137] It can be seen that the amplitude of the current signal I ripple and the voltage signal amplitude V ripple The following relationship exists:

[0138] I ripple =2πf·C·V ripple

[0139] The calculation formula of the capacitance value can be obtained as follows:

[0140]

[0141] E3. In actual applications, the system uses the voltage and current amplitudes obtained in step C and optimized in step D to calculate the capacitance at two frequency points: the grid fundamental frequency and the harmonic frequency. Specifically, for the two frequency points of 50 Hz and 100 Hz, the amplitudes are calculated using the synchronous demodulation and integral filtering methods described above. These amplitudes are then substituted into the capacitance calculation formula to obtain the capacitance values ​​at the two frequency points:

[0142] For the fundamental frequency point, that is, 50Hz frequency point, calculate the first capacitance value

[0143]

[0144] in, is the estimated value of the optimized current amplitude at the fundamental frequency point, is the estimated value of the optimized voltage amplitude at the fundamental frequency point.

[0145] For the harmonic frequency point, that is, 100Hz frequency point, calculate the second capacitance value

[0146]

[0147] in, is the estimated value of the optimized current amplitude at the harmonic frequency point, is the estimated value of the optimized voltage amplitude at the harmonic frequency point.

[0148] E4. Multi-frequency collaborative analysis. Due to various interference factors in the power system (such as harmonics, transient impacts, etc.), the capacitance value calculated at a single frequency point may have errors. The present invention adopts a multi-frequency collaborative analysis strategy, which is as follows:

[0149] Calculate the relative difference in capacitance between two frequency points:

[0150]

[0151] A preset threshold δ is set, for example, 5%.

[0152] When the relative difference ΔC < δ, the capacitance calculation results at the two frequency points are considered reliable, and the arithmetic mean is used as the final capacitance value:

[0153]

[0154] When the relative difference ΔC ≥ δ, it is necessary to determine which frequency point has a more reliable calculation result. The system makes this judgment by comparing the signal-to-noise ratio of the signals at the two frequency points and selects the capacitance value at the frequency point with a higher signal-to-noise ratio as the final capacitance value:

[0155]

[0156] E5. Capacitance evaluation. The system compares the calculated capacitance C with the rated capacitance C of the DC support capacitor. rated Compare and evaluate the health of the DC link capacitors.

[0157] Specifically, in one embodiment of the present invention, a health ratio can be calculated first: Health ratio =δ / C rated ×100%, then evaluate the health status according to the health ratio. ratio ≥90%, the health status is good; if 80%≤Health ratio <90%, the health status is normal; if 70%≤Health ratio <80%, then the health status is warned; if Health ratio If it is less than 70%, it needs to be replaced.

[0158] The online capacitor capacitance parameter identification method based on multi-frequency collaborative analysis provided by the present invention has the following technical effects:

[0159] High-Precision Measurement: By combining adaptive time-slot pulse weighting with synchronous demodulation, the target frequency amplitude extraction error is reduced to less than 1%, significantly improving online measurement accuracy. Compared to the 4.2%-6.8% error rate of traditional methods in complex electromagnetic environments, this method significantly improves the accuracy of capacitance measurement and provides a more reliable data foundation for DC link capacitor condition assessment.

[0160] Fast, real-time response: A simplified computational process based on discrete integration and pulse weighting enables real-time computation on an embedded platform, meeting the high responsiveness requirements of online monitoring. Compared to the response delay of over 500ms with traditional methods, this method shortens the response time to 30-50ms, significantly improving the system's ability to monitor changes in the state of DC link capacitors in real time, effectively supporting rapid early warning and resolution of grid faults.

[0161] Multi-frequency integrated correction: Simultaneously processes the fundamental frequency and harmonic frequency components, reducing the impact of external interference through multi-point data calculation, achieving more stable capacitance calculations. Especially in power grid environments with high harmonic content, multi-frequency collaborative analysis can effectively suppress the errors that may be caused by single-frequency calculations, improving the stability and reliability of capacitance identification, and providing a more comprehensive parameter basis for DC link capacitor health status assessment.

[0162] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for online identification of DC link capacitor capacitance based on multi-frequency collaborative analysis, characterized in that: include: S1. Using a sensor system to collect ripple voltage signals and ripple current signals at both ends of the DC link capacitor; S2. Determine a delay step length according to the grid fundamental frequency and the sampling frequency, and construct a pulse weight sequence based on the delay step length; S3. Performing a discrete convolution operation on the pulse weight sequence, the ripple voltage signal, and the ripple current signal to extract a periodic signal corresponding to the grid fundamental frequency and its harmonic components; S4. Process the periodic signal and extract the signal amplitudes at the grid fundamental frequency point and harmonic frequency point respectively; S5. Traverse the optimization parameters through the parameter search algorithm to determine the optimal parameter combination; S6. Re-execute steps S2 to S4 using the optimal parameter combination to obtain optimized amplitudes of the ripple voltage signal and the ripple current signal at the grid fundamental frequency point and harmonic frequency point; S7. Based on the voltage-current relationship of the capacitor, calculate a first capacitance value according to the amplitude relationship of the optimized ripple voltage signal and the ripple current signal at the grid fundamental frequency point, and calculate a second capacitance value according to the amplitude relationship of the optimized ripple voltage signal and the ripple current signal at the harmonic frequency point; S8. Perform multi-frequency collaborative analysis based on the first capacitance and the second capacitance to determine the actual capacitance of the DC link capacitor.

2. The method for online identification of DC link capacitor capacitance based on multi-frequency collaborative analysis according to claim 1, characterized in that: In step S2, the delay step length D is determined by: where f s is the sampling frequency, and f0 is the fundamental frequency.

3. The method for online identification of DC link capacitor capacitance based on multi-frequency collaborative analysis according to claim 1, characterized in that: In step S2, the construction formula of the pulse weight sequence h(n) is: Where δ(n) is the discrete unit pulse, M is the number of pulses, D is the delay step, and n is the discrete time index.

4. The method for online identification of DC link capacitor capacitance based on multi-frequency collaborative analysis according to claim 1, characterized in that: Step S1 specifically includes: The ripple voltage signal and the ripple current signal at both ends of the DC link capacitor are collected at a fixed sampling frequency; The data of the first N sampling points are selected for processing, where N is a preset value, to ensure that the selected data covers one or more complete cycles of the grid fundamental frequency.

5. The method for online identification of DC link capacitor capacitance based on multi-frequency collaborative analysis according to claim 1, characterized in that: In step S3, the discrete convolution operation performs time-domain convolution on the pulse weight sequence with the ripple voltage signal and the ripple current signal respectively, thereby achieving selective enhancement and noise suppression of specific frequency components, and obtaining weighted voltage signals and current signals.

6. The method for online identification of DC link capacitor capacitance based on multi-frequency collaborative analysis according to claim 1, characterized in that: In step S4, the signal processing includes multiplying the extracted periodic signal with the cosine function and the sine function of the corresponding frequency to obtain an in-phase component and an orthogonal component, and performing discrete integral filtering, amplitude compensation and tail-end averaging on the in-phase component and the orthogonal component.

7. The method for online identification of DC link capacitor capacitance based on multi-frequency collaborative analysis according to claim 6, characterized in that: The multiplication of the extracted periodic signal with the cosine function and the sine function of the corresponding frequency respectively means: for the grid fundamental frequency point and the harmonic frequency point respectively, the convolved voltage signal is multiplied with the cosine function of the corresponding frequency to obtain the voltage in-phase component, and multiplied with the sine function of the corresponding frequency to obtain the voltage orthogonal component; the convolved current signal is multiplied with the cosine function of the corresponding frequency to obtain the current in-phase component, and multiplied with the sine function of the corresponding frequency to obtain the current orthogonal component.

8. The method for online identification of DC link capacitor capacitance based on multi-frequency collaborative analysis according to claim 7, characterized in that: The discrete integral filtering process includes: Construct an integrating filter whose impulse response is of length L lp The uniform sequence, L lp is the integration window length; Performing convolution operations on the integral filter with the in-phase component and the quadrature component of the voltage and the current, respectively, to obtain filtered in-phase component and quadrature component of the voltage and the current; The amplitude compensation includes: Amplify and compensate the filtered in-phase component and quadrature component to correct the amplitude attenuation caused by integral filtering; The tail section averaging process includes: The voltage and current amplitudes at each time point are calculated separately, and the tail section of the signal is selected for averaging to obtain the final voltage and current amplitude estimates.

9. The method for online identification of DC link capacitor capacitance based on multi-frequency collaborative analysis according to claim 1, characterized in that: In step S5, the search space of the parameter search algorithm includes: the value range of the pulse number is 2 to 10, the value range of the integral filter length parameter is 1 to 5, and the value range of the tail segment average ratio parameter is 2 to 10.

10. The method for online identification of DC link capacitor capacitance based on multi-frequency collaborative analysis according to claim 1, characterized in that: The multi-frequency collaborative analysis in step S8 includes: When the difference between the first capacitance and the second capacitance is less than a preset deviation threshold, the arithmetic mean is used as the actual capacitance of the DC link capacitor; When the difference between the first capacitance and the second capacitance is greater than or equal to a preset deviation threshold, preferentially selecting one of the capacitances as the actual capacitance of the DC link capacitor according to the signal-to-noise ratio; The signal-to-noise ratio is determined by signal amplitude, spectrum characteristics or historical data statistical characteristics.

Images